文献速递牙周病口腔修复与种植

基于深度学习关键点检测的全口牙槽骨丧失定量自动化框架

An automated framework for quantitative alveolar bone loss using deep learning-based landmark detection.
PubMed 文献 · J Dent · 2026-03-10

本研究开发并评估了一种整合深度学习关键点检测与曲线拟合的全景片全口牙槽骨丧失(ABL)自动定量框架,使用760张全景片(532/152/76)由两名牙医标注并有第三专家裁定。TransPose关键点定位精度最高(MRE 12.50 px,SDR 80.77±5.79% at 15 px),YOLOv8与手动测量一致性最强(上颌ICC 0.633,下颌ICC 0.771),并在15%和33%阈值下分别达到敏感度/特异度0.87/0.71和0.57/0.98,自动分析仅需12.3 ms。

要点

  • TransPose在关键点定位中MRE最低(12.50 px),SDR最高(80.77±5.79% at 15 px)。
  • YOLOv8与手动测量的一致性最强(上颌ICC 0.633,下颌ICC 0.771)。
  • 自动分析仅需12.3 ms,远快于手动标注的14.2 min每张。

核心结论

  • 基于深度学习的框架可高效定量全景片上的牙槽骨丧失,TransPose定位精度最高,YOLOv8与临床评估最一致。

研究设计与样本

  • 共760张全景片按532/152/76划分,由两名训练有素的牙医标注,第三专家裁定。
  • 评估TransPose和HRNet热图关键点定位及YOLOv8单阶段检测器关键点回归,用MRE和SDR评价定位性能。
  • 通过多项式曲线拟合得出全口和区域ABL百分比,用ICC量化与手动测量的一致性,并在15%和33%阈值下评价筛查性能。

主要发现

  • TransPose的MRE最低为12.50 px(IQR 11.31-13.55),SDR最高为80.77±5.79%(15 px)。
  • YOLOv8与手动测量一致性最强,上颌ICC 0.633、下颌ICC 0.771。
  • 初步筛查中YOLOv8在15% cutoff敏感度/特异度0.87/0.71,在33% cutoff为0.57/0.98。
  • 手动标注每张图像需14.2 min,自动分析在12.3 ms内完成。

关键数值与推荐

  • 关键点定位评价指标为平均径向误差(MRE)和成功检测率(SDR),SDR在15 px阈值下报告。
  • ABL筛查性能在15%和33%阈值下评价,报告敏感度和特异度。
  • 与手动测量的一致性用组内相关系数(ICC)量化。

临床怎么用

  • 该AI工具可快速、客观地测量全景片上的ABL,有望支持临床实践中的标准化筛查。
  • YOLOv8与临床评估一致性较好,可能适合用于全口ABL的自动定量初筛。

局限性

  • 筛查性能评估为初步结果,需进一步验证。
  • YOLOv8与手动测量的ICC为中等水平(上颌0.633)。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

OBJECTIVE: To develop and evaluate an automated framework for full-mouth quantification of radiographic alveolar bone loss (ABL) on panoramic radiographs by integrating deep-learning landmark detection with curve fitting. METHODS: A total of 760 PANs (532/152/76) were annotated by two trained dentists, with third-expert adjudication. Three networks were evaluated: TransPose and HRNet for heatmap-based landmark localization, and YOLOv8 as a one-stage detector adapted for landmark regression. Landmark localization performance was evaluated using mean radial error (MRE) and success detection rate (SDR). Full-mouth and regional ABL percentages were derived via polynomial curve fitting, and agreement with manual measurements was quantified using intraclass correlation coefficients (ICC). Screening performance was evaluated at ABL thresholds of 15% and 33%. RESULTS: TransPose achieved the lowest MRE (12.50 px; IQR, 11.31-13.55) and the highest SDR (80.77 ± 5.79% at 15 px). YOLOv8 demonstrated the strongest agreement with manual measurements (ICC 0.633 for the maxilla; 0.771 for the mandible). In preliminary screening evaluation, YOLOv8 achieved sensitivity/specificity of 0.87/0.71 at the 15% cutoff and 0.57/0.98 at the 33% cutoff. Manual annotation required 14.2 min per image while automated analysis was completed within 12.3 ms. CONCLUSIONS: The AI-based approach enables efficient quantification of ABL using PANs. TransPose achieved the highest localization accuracy, whereas YOLOv8 best aligned with clinical evaluations. CLINICAL SIGNIFICANCE: This AI tool offers a rapid, objective alternative for ABL measurement on PANs, potentially supporting standardized screening in clinical practice.

来源:J Dent,查看原文。版权归原作者所有。

引用

  1. Tian E, Huang L, Fu B, et al. An automated framework for quantitative alveolar bone loss using deep learning-based landmark detection. J Dent. 2026. PMID: 41812977. doi:10.1016/j.jdent.2026.106620

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。